About LimitedPolicy

HoraeDBClient will limit the write/query by default (the default maximum number for write requests is 16384, and the default maximum number of requests for queries is 8), The limiting parameters see the configuration in WriteOptions / QueryOptions.

LimitedPolicy is the processing policy after the flow is limited. It should be noted that if the number of data contained in one write request exceeds the limit, then the request will be allowed.

name	description
DiscardPolicy	The latest written data is directly discarded when limiting
AbortPolicy	Throw a LimitedException directly to the upper layer when limiting
BlockingPolicy	Blocking thread while limiting
BlockingTimeoutPolicy	Blocking thread util timeout，the error message while be reflected in Result#Err when timeout occurs, and no exception will be thrown
AbortOnBlockingTimeoutPolicy	Blocking thread util timeout，and throw a LimitedException to the upper when timeout occurs
XXXPolicy	If none of the above meets your needs, please customize it based on LimitedPolicy. Happy coding : )

Note：default write limit policy is AbortOnBlockingTimeoutPolicy(3s)，query is AbortOnBlockingTimeoutPolicy(10s)

Adaptive limiting

HoraeDBClient also provides an adaptive current limiter based on TCP Vegas and Gradient limit algorithms at the underlying communication layer.

Vegas

Vegas is a congestion control algorithm that actively adjusts cwnd. The main idea is to set two thresholds, alpha and beta, and then calculate the target rate (Expected) and the actual rate (Actual), then compare the relationship between diff and alpha and beta, and adjust cwnd.

The core algorithm is

diff = cwnd*(1-baseRTT/RTT)
if (diff < alpha)
    set: cwnd = cwnd + 1
else if (diff >= beta)
    set: cwnd = cwnd - 1
else
    set: cwnd = cwnd

Gradient Concurrency-limits

The Gradient concurrency-limit algorithm adjusts the limit according to the change gradient of the current average RTT and the long-term exponentially smoothed average RTT. Unlike traditional congestion control algorithms, average values are used instead of minimum values.

The core algorithm is

// Compute the gradient limited to the range [0.5, 1.0] to filter outliers
gradient = max(0.5, min(1.0, longtermRtt / currentRtt));

// Calculate the new limit by applying the gradient and allowing some queuing
newLimit = gradient * currentLimit + queueSize;

// Update limits with smoothing factor (default is 0.2)
newLimit = currentLimit * (1 - smoothing) + newLimit * smoothing

• The default adaptive current limiting algorithm is Gradient，you can set in configuration org.apache.horaedb.rpc.RpcOptions.limitKind = Vegas by Vegas, or set org.apache.horaedb.rpc.RpcOptions.limitKind = None to close
• It will fail quickly when the limit is reached by default, and it can also be set by setting org.apache.horaedb.rpc.RpcOptions.blockOnLimit = true
• Write and read requests share this current limiter. By default, 70% and 30% of the access permissions can be obtained for write and read respectively. You can use the startup parameters -DCeresDB.rpc.write.limit_percent=xxx，readLimitPercent is 1 - writeLimitPercent